A multi-layered modelling and simulation architecture Julie Acquaviva*, Jean-Baptiste Filippi, Paul Bisgambiglia UMR CNRS 6134, University of Corsica, Quartier Grossetti 20250 Corte, France {jacquaviva, filippi, bisgambi}@univ-corse.fr Abstract: Modelling and simulation have taken a preponderant place in the analysis and design of complex systems. The general modelling and simulation environment developed at the UMR CNRS 6134 lab in recent years enables the study of natural complex systems. However, such studies are often tedious and the approach we present here intend to make easier the modelling and simulation of complex spatially distributed systems. This multi-layered approach enables an effective tasks distribution among different specialists, and makes the use of existing models easier. The multi-layered approach is introduced after a brief review of the background of our method. The method is illustrated by the analysis of one complex natural system, a watershed. Keywords: Discrete Event Simulation, Multi-Layered Modelling, Watershed Model, Environmental Modelling 1 INTRODUCTION The research in defining and creating a general purpose modelling and simulation environment began ten years ago at the UMR CNRS 6134 lab of the University of Corsica. This research has resulted in the definition of the JDEVS modelling and simulation environment [Filippi et al., 2004]. This experimental environment might be used to help modelling complex problems solvable by discrete-event simulation and is especially suited for natural system modelling and simulation. Nevertheless, natural systems modelling calls for many subjects and integrating all these disciplines can turn out to be tedious when the model has properties that evolve in space. A new modelling methodology has been conceived to avoid the models redefinition and to allow a more efficient tasks distribution. The multi-layered approach is based on an automatic coupling of spatially distributed models. With this methodology, the system is represented by thematic layers, each of them representing a specific field of the studied system. A review of the basic concepts of the DEVS formalism and the JDEVS modelling and simulation environment will be given in the first part of this paper. The new approach for spatial models coupling is presented and discussed in the second part of the paper followed by an application of this method in the last section. 2 THE MODELLING ENVIRONMENT This section presents the mathematical foundations of our approach and the software environment that has been developed upon. A review of the DEVS formalism Zeigler developed the discrete event system specification (DEVS) formalism as a mathematical basis for discrete event modelling [Zeigler, 1984], [Zeigler, 1990]. Applying this formalism, models are constructed in a modular and hierarchical manner. DEVS formalism is based on the use of two kinds of models: atomic and coupled models. Atomic models describe the functionality of basic system entities. Coupled models correspond to more complex entities. The inner structure of the complex entities is represented by the components of the coupled model and their coupling. Both atomic and coupled models can be used as model components. An atomic model is defined by the 7-tuple: AM = where X = set of external input events; S = set of sequential states; Y = set of output events; dint: S ? S = internal transition function; dext: QXX ? S = external transition function; where Q = {(s,e)|s ? S, 0=e=ta(S)} = total state set; ?: S ? Y = output function; ta: S ? R0 + = time advance function. Coupled models are defined in DEVS formalism by the 7-tuple: CM = where .....